That's a group effort. For picture they could read metadata for something as simple as Paris. The tech is there to search for 'architecture' pics, Google Phot

That's a group effort. For picture they could read metadata for something as simple as Paris. The tech is there to search for 'architecture' pics, Google Photos demonstrates it. Of course there is the issue of who can see this data and that's not a small one. The recognition itself could run locally, but training data would still need to be mutualized. And most apps would need to adapt and provide more metadata to your documents